Confirmations and finality: when is a transaction actually done?
Some chains give you certainty in seconds, others give you steadily improving odds. Knowing which kind you are using decides how long you should wait before treating a payment as settled.
Reviewed by Lauren Bennett, Senior Bitcoin Analyst · Last reviewed October 5, 2026
On this page
A confirmation is one block built on top of the block containing your transaction. Finality is the point at which reversing that transaction becomes impossible rather than merely expensive. Proof-of-work chains offer probabilistic finality that strengthens with every block; many proof-of-stake chains offer explicit finality after a defined interval, backed by stake that would be destroyed if validators changed their minds.
Why is one block not enough?
Blocks are produced by independent participants who do not learn about each other's work instantly. Two valid blocks can be produced at nearly the same height, and the network resolves the conflict by continuing to build on one of them. The losing branch is discarded, and any transaction that appeared only there returns to the pending pool.
That reorganisation — a reorg — is normal. It is why a single confirmation is a weak guarantee, and why the strength of a guarantee grows with the amount of work or stake committed on top of it.
How does probabilistic finality work?
On a proof-of-work chain, rewriting history means redoing the work of every block since the one you want to change, faster than the honest network extends the chain. Each additional confirmation multiplies that cost. There is no threshold at which reversal becomes impossible — only a depth at which it becomes economically absurd.
This is the reasoning behind exchange deposit thresholds. A small transfer might be credited after a couple of confirmations; a very large one may wait considerably longer, because the value at stake changes the calculation. Our guide to proof of work versus proof of stake covers the security models behind both approaches.
What does finality mean on proof-of-stake chains?
Many proof-of-stake designs add an explicit finalisation step: validators vote, and once a supermajority has attested across a defined interval, the block is finalised. Reverting it would require those validators to have signed conflicting statements, which the protocol punishes by destroying their stake. The guarantee is therefore economic and explicit rather than statistical.
| Model | What accumulates | Practical wait |
|---|---|---|
| Proof of work | Cumulative work on top | Several confirmations, scaled to value |
| Proof of stake with finality | Validator attestations, then finalisation | Minutes, then effectively irreversible |
| Fast-finality chains | Consensus among a smaller validator set | Seconds, with different trust assumptions |
| Layer-2 rollup | Sequencer confirmation, then base-layer settlement | Instant locally; base-layer finality later |
What actually causes a reorg?
- Two blocks found at nearly the same time — routine and usually resolved within one block.
- Network partitions, where parts of the network build separately before reconnecting.
- Client bugs or upgrade mismatches, which can split the network along software lines.
- Deliberate attacks, which require overwhelming resources and are rare on major chains.
For an ordinary user, the practical implication is narrow: do not treat a single confirmation as settlement for anything valuable, and be sceptical of a service that credits large amounts instantly on a chain with probabilistic finality.
How long should you wait?
Match the wait to what you would lose if the transaction were reversed. For coffee, one confirmation is fine. For a house-sized transfer, wait for the depth the receiving institution requires and then some. If you are the one receiving funds from a stranger, use the receiving platform's threshold rather than your own patience — those numbers exist because someone modelled the risk.
You can watch confirmations accumulate on any explorer; our guide to using a block explorer covers where the count appears and what else to check while you wait.
What does this mean for merchants and payments?
Anyone accepting crypto as payment faces the same trade-off as an exchange, with less tolerance for delay. A coffee shop cannot ask a customer to wait twenty minutes, and does not need to: the amount at risk is small enough that a zero-confirmation acceptance is a rational business decision, priced like any other fraud loss.
As the amount rises the calculation flips. For a car or a property, the sensible policy is to wait for the depth at which reversal would require resources nobody plausibly has, and to say so up front so the wait is expected rather than alarming. Payment processors formalise exactly this by setting thresholds per amount and absorbing the residual risk themselves.
- Small amounts: accept quickly, treat the residual risk as a cost of doing business.
- Medium amounts: wait for the standard threshold on that chain.
- Large amounts: wait for finality where the chain offers it, or for substantial depth where it does not.
- Any amount from an unknown counterparty: apply the higher threshold regardless of size.
- 1. Proof of stake and finality — ethereum.org
- 2. Chain reorganisations and forks — ethereum.org
- 3. Bitcoin: A Peer-to-Peer Electronic Cash System (section 11) — Satoshi Nakamoto
Frequently asked
What is a blockchain confirmation?
One block added on top of the block containing your transaction. Each one increases the work or stake that would have to be undone to reverse it.
What is the difference between finality and confirmations?
Confirmations measure accumulated certainty on chains where reversal is always theoretically possible. Finality is an explicit protocol guarantee that a block cannot be reverted without validators losing their stake.
Are reorgs dangerous?
Short reorgs are routine on some networks and affect only very recent blocks. They matter if you treat a single confirmation as settlement, which is why thresholds exist.
Why do exchanges require different confirmation counts?
Because the security models and block times differ per chain, and because the amount at risk changes the calculation. Their thresholds reflect their own risk modelling.
Is a layer-2 transaction final immediately?
It is confirmed by the rollup quickly, but full base-layer settlement comes later when its data is posted and, for some designs, after a challenge window.
Mason Walker is an Ethereum & Layer-2 Editor at Crypto Almanac Daily, where he covers the rapidly evolving Ethereum ecosystem with a focus on staking, Layer-2 networks, rollups, protocol upgrades, and smart contract infrastructure. His reporting explores how Ethereum's technical innovations shape decentralized finance, tokenization, and Web3 applications. Mason specializes in breaking down complex protocol changes, network scalability solutions, validator economics, and the growing adoption of optimistic and zero-knowledge rollups into accessible, research-driven analysis. Before joining Crypto Almanac Daily, Mason covered blockchain infrastructure and emerging financial technologies, developing expertise in Ethereum's architecture and the broader smart contract ecosystem. His work combines technical accuracy with clear explanations, helping readers understand both the engineering behind blockchain networks and their real-world market implications. At Crypto Almanac Daily, Mason writes daily news, protocol deep dives, ecosystem updates, educational guides, and long-form research articles, providing readers with reliable insights into Ethereum's ongoing development and its role in the future of decentralized finance and digital assets.
This guide is educational and general in nature. It is not financial, investment, legal or tax advice, and it does not account for your circumstances. Crypto assets are volatile and you can lose the money you put in. See our editorial policy and methodology.
